About this podcast
Chris Paxton & Michael Cho geek out over robotic papers with paper authors.
robopapers.substack.com
Episodes
The newest 20 of 113 episodes
- Ep#114: Oopsie Data · 44 minRobotics has a data problem, but not the one you’re thinking of: almost all robotics data is of successful instances of robot behaviors. But many of the most exciting research…
- Ep#113: MolmoBOT: Large-Scale Simulation Enables Zero-Shot Manipulation · 39 minUsing simulation to teach robots would unblock the massive scale of robotics data that seems to be missing right now. The issue is usually sim-to-real transfer; MolmoBOT aims to…
- Ep#112: SlipSense and Touch2Trace · 1 h 2 minTouch is crucial for humans when they want to perform precise tasks to a high degree of reliability, and yet robots almost universally use vision — they’re often completely blind…
- Ep#111: SOLE-R1: Video-Language Reasoning as the Sole Reward for On-Robot Reinforcement Learning · 53 minReinforcement learning is often seen as key to achieving scalable, highly reliable real-world robot performance, yet it requires good reward functions that can supervise the wide…
- Ep#110: ENPIRE: Agentic Robot PolicySelf-Improvement in the Real World · 54 minIn many fields, automated research has become increasingly plausible with the improving capabilities of frontier language models, and the field robotics may be no exception. In…
- Ep#109: Enigma - Connecting 100 Robots To The Internet · 49 minWhat happens if you connect a hundred AI-powered robots to the internet and just let people give them any instructions they want? Enigma founder Jonathan Jacobi joins us to talk…
- Ep#108: VLS: Steering Pretrained Robot Policies via Vision–Language Models · 52 minPretrained vision-language-action models for robots often fail when a task is performed in a slightly different environment. With Vision-Language Steering, you can get better…
- Ep#107: Do As I Do: Dexterous Manipulation Data from Everyday Human Videos · 1 h 2 minAchieving human-like dexterity is the next frontier for robotics, and yet dexterity data is often subtly hard to scale. Real-world dexterity data, including things like…
- Ep#106: Flex-π: A Multi-Stream World-Action Model with Compute Flexibility · 54 minWorld Action Models are becoming more popular in robotics, as they learn to predict the world jointly with learning how to act on it. However, these world predictions are usually…
- Ep#105: Beyond Imitation: Self-Improving Robot Policies via Off-Policy Q-Planning · 50 minImitation learning, especially with interventions, has driven so much recent robotics progress. However, improving a policy via targeted interventions until it reaches a useful…
- Ep#104: Light-Loco-Parkour: Versatile Perceptive Whole-Body Locomotion via Multi-Skill Distillation · 43 minOne of the key advantages of legged robots like humanoids should be how effectively they can move across a wide variety of terrain types to accomplish their task. But…
- Ep#103: Freeform Preference Learning for Robotic Manipulation · 48 minSpecifying reward functions for robots is one of the hardest things about reinforcement learning. Robot rewards often need to be very detailed; metrics like progress can be…
- Ep#102: EgoSuite-Open100K from Lightwheel · 1 h 9 minRobotics has a data problem, and egocentric data is a compelling way to solve it because egocentric data is (1) scalable — it’s cheap to collect — and (2) easily captures the…
- Ep#101: Coupled Local and Global World Models for Efficient First Order RL · 51 minReinforcement learning has unlocked tremendous gains in broader AI and humanoid robot whole body control, but still has yet to deliver strong results for robot manipulation due…
- Ep#100: 100 Episodes of Embodied AI · 1 h 29 minEmbodied AI is exploding, and it’s been amazing to learn about it with you all. We have now recorded over a hundred discussions with the best researchers in the world on embodied…
- Ep#99: DYNA-2: A 1 Million Hour Scaling Law for World-Action Models · 1 h 7 minExcitement about world-action models and robot learning has never been higher — they promise a way to use human egocentric data to train massive robotics models which can provide…
- Ep#97: Cross-Embodiment Robot Manipulation via a Unified Hand Action Space · 60 minRepresenting actions for robot hands is actually very challenging. Robot hands have different sizes and shapes, and different degrees of freedom. Representing their actions…
- Ep#96: EgoHumanoid: Unlocking In-the-Wild Loco-Manipulation with Robot-Free Egocentric Demonstration · 37 minHow can roboticists scale human manipulation data in the wild? EgoHumanoid builds a comprehensive pipeline which unlocks mobile manipulation training for robots — meaning that…
- Ep#95: Action-to-Action Flow Matching · 49 minDiffusion Policy was one of the big breakthroughs that has enabled an explosion in real-world robot learning. However, it’s always had a weakness, which is that it works by…
- Ep#94: Learning a Thousand Tasks in a Day · 1 h 16 minFor general-purpose robots to be useful, we must be able to teach them new tasks quickly and repurpose them to new roles. But most techniques for teaching robots — even from…
